Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

goMac

macrumors 604
Original poster
Apr 15, 2004
7,663
1,694
My Mac Pro 5,1 has been having problems for a bit. A few months about, I started getting gpuRestarts and random lockups on my XFX Radeon 580. I thought maybe the GPU was bad, so I switched to the Sapphire Pulse 580, and am having the same issues.

I was wondering if anyone else had seen failure of a slot. I moved the GPU out of slot one, which is not ideal, but so far seems stable. The weird thing is after a crash, my GPUs link speed will drop to 2.5 GT/s. Usually when restarting I'll immediately get another few lockups and panics before it goes away a bit, which is also making me think hardware and not software.

Has anyone seen a PCIe slot fail on a Mac Pro 5,1? The only weird thing is if I drop in a Geforce 680, everything is working fine. The 680 does seem to be a cooler card, so I'm not sure if the problem could be heat related or how that would relate to slot 1 being the issue.

I'll run the GPU in slot 2 for a few days and see if the problem comes back. Just wondering if anyone has seen anything similar.
 
What OS version and firmware version?

I’ve had problems with my system and the Sapphire RX580, however the GPU finally failed to output a video signal. The expectation is that my issues were caused by a defective GPU (System is fine with HD7970), but I won’t know for sure until my RMA replacement GPU arrives.

Keep in mind that sometimes similar symptoms are caused by different problems.

BTW, I put the GPU in PCIe slot 2 for better cooling.
 
What OS version and firmware version?

I’ve had problems with my system and the Sapphire RX580, however the GPU finally failed to output a video signal. The expectation is that my issues were caused by a defective GPU (System is fine with HD7970), but I won’t know for sure until my RMA replacement GPU arrives.

Keep in mind that sometimes similar symptoms are caused by different problems.

BTW, I put the GPU in PCIe slot 2 for better cooling.

I believed slot 2 provide better cooling as well because the air can flow through both upper and lower side.

However, in my test, the card actually has better cooling in slot 1. Not sure why, may be because warm air rise, so the higher the slot, the warm the air the graphic card will intake, but this is the test result (the test was done with a PNY XL8 1080Ti).

Or may be because all 4 HDD bays are occupied. Therefore, too little space above the card for the warmed air to escape.
 
I believed slot 2 provide better cooling as well because the air can flow through both upper and lower side.

However, in my test, the card actually has better cooling in slot 1. Not sure why, may be because warm air rise, so the higher the slot, the warm the air the graphic card will intake, but this is the test result (the test was done with a PNY XL8 1080Ti).

Or may be because all 4 HDD bays are occupied. Therefore, too little space above the card for the warmed air to escape.

In my case, I have two long x16 cards. If I put the GPU in slot 1, my 7107A blocks the fan of my HD7970 and the GPU overheats. It gets so hot it trips the resettable fuse, and I can’t use the GPU for 48 Hours.

This doesn’t happen at all if GPU is in slot 2. Behavior is consistently reliable..

With RX580, GPU is so big, I can’t use 7101A if GPU is in slot 1. It physically interferes with 7101A in Slot 2.
 
What OS version and firmware version?

10.14.3, but I've been having this problem through the 10.14.2 betas including the GM. Boot ROM 140.0.0.0.0.

If you dont need the Slots theres nothing bad about using Slot 2. Slots 3 & 4 are slower 4x slots.

I have an SSD in slot 4 that partially covers the fan, so I'm not totally excited about using slot 2. If I trusted slot 1 I could move the SSD there I guess.

I believed slot 2 provide better cooling as well because the air can flow through both upper and lower side.

So here's the really weird thing that's important that it looks like I left out of my initial post...

The card only gets unstable when it's cool.

I can run it at 70 C all day just fine no issues. 8 hour long stretches and everything is great. But every once in a while when it cools down to low 40s/mid 30s the card starts crashing in slot 1. And the problems don't clear up until a few restarts.

If I switch the card to the low power BIOS the issue gets even worse.

The 580 puts out so much ambient heat I'm wondering if something on the board is flexing as it cools that's not supposed to be. Maybe a pin in the connector is misaligned just enough. But in searching around I can't find anyone who's ever had a similar issue.
 
10.14.3, but I've been having this problem through the 10.14.2 betas including the GM. Boot ROM 140.0.0.0.0.

I have an SSD in slot 4 that partially covers the fan, so I'm not totally excited about using slot 2. If I trusted slot 1 I could move the SSD there I guess.

So here's the really weird thing that's important that it looks like I left out of my initial post...

The card only gets unstable when it's cool.

I can run it at 70 C all day just fine no issues. 8 hour long stretches and everything is great. But every once in a while when it cools down to low 40s/mid 30s the card starts crashing in slot 1. And the problems don't clear up until a few restarts.

If I switch the card to the low power BIOS the issue gets even worse.

The 580 puts out so much ambient heat I'm wondering if something on the board is flexing as it cools that's not supposed to be. Maybe a pin in the connector is misaligned just enough. But in searching around I can't find anyone who's ever had a similar issue.

AFAIK, 2D and 3D use different part of the GPU, may be only the 2D part is faulty (assume move the card to slot 2 does NOT fix the issue).

However, if change slot fix the problem, then give the PCIe slot 1 a good blow. A little piece of dust in the slot can cause this kind of instability.
[doublepost=1546336078][/doublepost]
In my case, I have two long x16 cards. If I put the GPU in slot 1, my 7107A blocks the fan of my HD7970 and the GPU overheats. It gets so hot it trips the resettable fuse, and I can’t use the GPU for 48 Hours.

This doesn’t happen at all if GPU is in slot 2. Behavior is consistently reliable..

With RX580, GPU is so big, I can’t use 7101A if GPU is in slot 1. It physically interferes with 7101A in Slot 2.

That make sense in your setup.

I used to have dual HD7950 as well. And if I install them in slot 1 and 2, the card in slot 1 will run very hot. So I have to move the 2nd card to slot 3 due to the same reason.
 
  • Like
Reactions: crjackson2134
AFAIK, 2D and 3D use different part of the GPU, may be only the 2D part is faulty (assume move the card to slot 2 does NOT fix the issue).

Mentioned in the original post I've tried two different 580s (XFX and Sapphire Pulse) and both have the problem. And so far everything is good in slot 2.
 
  • Like
Reactions: h9826790
I was getting random restarts after first getting my Sapphire PULSE RX 580, but I swapped out my PSU and haven't had a single issue since.
 
I was going to say this sounds like a power supply issue after reading the comments.
However, if a failure shows up when the GPU is cold and not hot, that sounds like a cold solder joint with a poor electrical connection when cold.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.